The short answer: Claude Fable 5.1 (model ID claude-fable-5-1) is live on HopBase; check GET /v1/models for your key, and existing keys only need the model name changed. It is called through the native Anthropic Messages API at /v1/messages, so the Anthropic SDK and Claude Code work with zero code changes. It has a 1M context window and 128K max output, and adaptive thinking is always on: you steer reasoning depth with output_config.effort instead of a thinking parameter. The list price matches Fable 5 at $10 input and $50 output per million tokens, with cache writes at $12.50 (5 min) or $20 (1 hour), but cache reads drop to $0.25 from $1.00, so long sessions and agent loops that re-read the same prefix save the most. When migrating, note three changes: tool_choice any or tool returns 400, thinking blocks must be passed back unchanged with append-only history, and assistant prefill is not supported.
Anthropic's newest flagship, Claude Fable 5.1 (model ID claude-fable-5-1), is live on HopBase. Check GET /v1/models for your key — existing keys only need the model name changed.
Model specs
- 1M context window, 128K max output: whole repositories, long documents, and long multi-turn conversations fit in a single request, with no need to split long outputs.
-
Adaptive thinking, always on: no
thinkingparameter needed; steer reasoning depth and spend withoutput_config.effort(low through max). - Built for long-horizon agent work: stronger than Fable 5 at long-running autonomous coding, document / spreadsheet / slide work, multi-step research, vision, and computer use. Same tokenizer as Fable 5, so token counts are essentially unchanged.
-
Native Anthropic Messages API: standard
/v1/messages; the Anthropic SDK and Claude Code switch over with zero code changes.
Pricing: official list price
| Item (per million tokens) | Claude Fable 5.1 | Claude Fable 5 |
|---|---|---|
| Input | $10 | $10 |
| Output | $50 | $50 |
| Cache write (5 min / 1 hour) | $12.50 / $20 | $12.50 / $20 |
| Cache read (hit) | $0.25 (0.025×) | $1.00 |
Cache reads cost a quarter of Fable 5's, so long sessions and agent loops that re-read the same prefix save the most. Billing is list price × actual usage, visible in the console in real time; for enterprise volume, request a quote from the pricing page.
Three things to check when migrating from Fable 5 / Opus 5
-
Forced tool use is gone:
tool_choiceany/toolreturns 400. Useautoplus an instruction naming the tool, andstrict: trueor structured outputs when you need schema-valid JSON. - Thinking blocks are bound to the model: pass them back unchanged within a conversation and keep the history append-only — editing earlier turns invalidates earlier thinking blocks.
- No assistant prefill: as with every Claude model since 4.6, control output format with structured outputs or the system prompt.
Three-minute quickstart
curl https://api.hop-base.com/v1/messages \
-H "x-api-key: sk-your-key" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-fable-5-1",
"max_tokens": 1024,
"messages": [{"role": "user", "content": "Hello, introduce yourself"}]
}'
Full setup steps for Claude Code and the Anthropic SDK are in the Anthropic protocol guide. If you already have a key, just switch the model name to claude-fable-5-1.
Top comments (0)